Internet (Internet) traffic is always in a rapid growth, and the growth of the traffic pushes network devices such as routers to continuously develop towards high-capacity. A previous router is of centralized processing, and packet forwarding is performed by one processor, but processing capability of the centralized router is limited. The emergence of a bus type distributed router increases the processing capability of the router, the bus type distributed router has multiple insertion cards, each insertion card has a processor that may process packets, the insertion cards are connected through a bus, an insertion card applies to use the bus when needing to send data to another insertion card, and the bus is dedicated during the period from the time a system agrees to completion of the sending, during which the other insertion cards cannot send data. The processing capability of the bus type distributed router is the total of the processing capability of all the insertion cards, a bandwidth of the bus is the bandwidth of the system. Actual products of the centralized and bus type distributed routers are usually single-frame devices.
The emergence of a switching network enables multiple pairs of insertion cards to communicate with one another at the same time, thus further increasing the capacity of the system. The capacity of the single-frame device is increased greatly through the architecture combining the switching network and the distributed processing, but on the whole, the growth speed of the traffic is higher than the development speed of the single-frame capacity, and the capacity of the single-frame device still cannot satisfy the demand. Therefore, the capacity of a network node needs to be increased through multi-frame or multiple devices.
One method to increase the capacity of a router node is networking by using multiple devices, interconnecting the devices by using standard lines, and replacing a single apparatus with a network. FIG. 1 is a router scheme of interconnecting by using multiple devices. As shown in FIG. 1, a node is set up by using four same routers (A, B, C, D), and each router has six standard 10GE interfaces. For example, the six interfaces of the router A are A1, A2, A3, A4, A5, and A6. The interfaces of the four devices are connected for networking, that is, the A4 interface of the router A is connected to a B4 interface of the router B, A5 is connected to an interface D5 of the router D, A6 is connected to an interface C6 of the router C, . . . , each device has three interfaces used for internal interconnecting (such as the interfaces A4, A5, A6 of the router A), the other three interfaces are used to provide a bandwidth to the outside (such as the interface A1, A2, A3 of the router A), so that the four devices each with a capacity of 60 G form a non-blocking network with a capacity of 120 G, which increases the capacity of the node.
The main disadvantages of the scheme of networking by using multiple devices in FIG. 1 are as follows: 1. Multiple devices lead to complexity of the network and a high management cost. 2. A packet needs to undergo packet processing for multiple times when passing through the network, for example, a packet, entering the network through the interface A1 and exiting the network through C1, at least needs to undergo packet processing performed by two routers, A and C, the work is repeated, the cost is high, and the power consumption is high.
Another method to increase the capacity of a router node is setting up a cluster router by using a dedicated central switching frame. FIG. 2 is a router clustering scheme of a central switching frame type. As shown in FIG. 2, one or more dedicated central switching frames are connected to multiple forwarding frames by using private interconnection lines, and the forwarding frames are responsible for packet forwarding processing. For example, a forwarding frame 1 receives a packet to perform forwarding processing, and decides to send the packet out through a forwarding frame 3, so that the forwarding frame 1 requests the central switching frame to transmit the packet to the forwarding frame 3, and the central switching frame switches the packet between the forwarding frames according to the request of the forwarding frame.
The disadvantages of the central switching frame type clustering in FIG. 2 are as follows: 1. The system is complex, implementation of a switching network is complex, multi-stage switching is usually required, extension is not flexible, and many devices only have single-stage switching for cost reasons and cannot adopt the clustering manner. 2. The extension capability of the switching network is limited, and the supported number of the forwarding frames and the supported number of line cards are limited, so that the capacity of the node is still limited. 3. The switching network is highly firmwared, which results in poor compatibility, and the switching network cannot be compatible with a conventional forwarding frame when updated and upgraded because an interconnection protocol between the switching frame and the forwarding frame is upgraded, thus failing to protect investments of users.